Build with DataHub: The Agent Hackathon

Editorial card, not a replacement for official terms. The text below is compiled from structured catalog information and explains how to assess the possibility of participation from Kazakhstan. Confirmed facts are clearly separated from analysis and practical recommendations. If the official website or legal rules conflict with this card, the latest version of the organizer's documents always takes precedence.

Brief profile

Theme: AI agents, data and ML. The standard exclusion list does not contain Kazakhstan.

Format and location: Online; Online

Period: 06.07.2026 — 10.08.2026

Registration or submission deadline: 10.08.2026

Current status: Open

Kazakhstan eligibility: Yes — according to the rules

For whom: AI/data developers, teams, and organizations

Age and other criteria: 18+ / age of majority

Team: Limit not specified

Main language: EN

Fee: No purchase required

Prizes: $20,500 cash; grand prize $6,000

What is this opportunity

Build with DataHub: The Agent Hackathon — an event in the "artificial intelligence, agents, and data" track. The recorded thematic formulation is: AI agents, data and ML. The standard exclusion list does not contain Kazakhstan.. The practical value of participation should be evaluated not only by the prize size, but also by the quality of the task, access to experts, suitability of the result for a portfolio, cost of preparation, and the likelihood of completing a strong submission in the available time. The catalog status is Open, and the editorial priority is A — apply now. This helps determine the order of work, but is not an organizer rating and does not guarantee selection or victory.

For this theme, the central product question is: which user decision is improved with the help of the model and why a regular deterministic process is not enough. A strong team will be able to answer this in one sentence and then prove the answer with a working result. The most appropriate type of final artifact: a working vertical scenario, a set of test cases, an evaluation log, an explanation of data and model, and safe failure handling. This is an editorial recommendation derived from the theme; the official mandatory format should only be taken from the event website.

Participation of a team from Kazakhstan

In the card, eligibility is marked as Yes — according to the rules. The recorded basis is: The standard exclusion list does not contain Kazakhstan.. These two lines must be read together. The phrasing "Kazakhstan is not excluded" is weaker than a direct invitation to participants from all over the world, and the word "international" by itself does not always mean the absence of restrictions on citizenship, residency, place of study, age, or sanctions compliance. If the value contains "conditionally", if the proof is indirect, or if the legal rules have not yet been published, it is worth getting a written response from the organizer before spending time.

The recommended inquiry to the organizer should state the country of residence and citizenship, the proposed team composition, age or student status, the method of receiving the prize, and the chosen format of participation. It is best to save the response along with a screenshot of the rules version. Separately check whether the platform applies its own restrictions, whether a Kazakhstani bank can receive the payment, whether a tax form is required, and whether participation on behalf of a company is allowed. These are not additional rules of this hackathon, but a standard risk check for an international application.

Who is it suitable for and how to assemble a team

Target audience: AI/data developers, teams and organizations. Fixed criteria: 18+ / age of majority. Team composition info: Limit not specified. Before registration, each participant must independently confirm compliance with age, student status, residency, and other personal conditions. If permission for solo participation is not explicitly stated, it should not be assumed to be allowed. If the maximum team size is unknown, do not register extra people before getting a response from the organizer.

For a task in the field of "artificial intelligence, agents and data", a rational distribution of roles looks like this: ML/agent engineer, product/backend developer, UX/domain researcher, and eval/demo lead. One person can combine several functions, but there must be clear owners of the product, technical result, quality assurance, and final submission. It is useful to agree in writing beforehand on the contribution, the right to represent the project, access to the repository, use of the result after the event, and distribution of the potential prize. Such an internal agreement is a recommendation and does not replace the platform's terms.

Working language is specified as EN. Even if the team communicates in Russian or Kazakh, prepare a unified glossary of terms in advance and a person who can confidently answer the jury's questions in the required language. For English submissions, it is better to use short sentences, captions on diagrams, and subtitles for the demo. The goal is accuracy, not a complex style.

Format, calendar and logistics

Fixed format — Online, location — Online. Start: 06.07.2026; end: 10.08.2026; deadline: 10.08.2026. Travel information: Not required. Presence of online component on the card: yes; signs of mandatory in-person presence: no. These signs are obtained only from the text of the fields and do not override event-specific rules.

For the online format, it is important for a team from Kazakhstan to convert the deadline to their time zone, check the time of mandatory sessions, the upload speed of large videos, and the availability of all APIs. Create a technical buffer of at least a few hours and do not leave registration until the moment of submission. For in-person or hybrid formats, first calculate the full budget: visa, flight, accommodation, insurance, city transport, meals, and refundable deposits. The phrase "travel support" does not mean automatic coverage of all costs. The fixed value for this event: Not required.

Fee, prizes and real value

Participation cost is recorded as No mandatory purchase. Prize information: $20,500 cash; grand prize $6,000. Before making a decision, separate cash payouts, grants, cloud platform credits, subscriptions, equipment, mentorship, and marketing statements about aggregate value. Clarify the number of winning teams, currency, deadlines, taxes, restrictions on receipt, and the need to attend the ceremony. If TBA, page discrepancy, or a re-check requirement is found in the field, the amount cannot be used as guaranteed income.

Even without a cash prize, the event can be useful if it provides a strong case study, public demonstration, feedback, or access to the community. But paid services, travel, and several weeks of work have an opportunity cost. The decision to participate is best made based on three questions: can the mandatory requirements be met; can a convincing result be assembled on time; will the created artifact be useful after the winners are announced?

Recommended project strategy

For this direction, the following sequence is useful:

Research. first define the task and the baseline non-AI process, collect a small representative eval set, then choose the minimal model and tools. Do not start with a list of technologies: first define the user, the problem, the baseline process, and the observable success criterion.
Project scope. Choose one end-to-end scenario that can be run from start to finish. Write down the features deliberately left out of the hackathon version.
Proof. Key metrics: quality on a fixed set, success rate of completed tasks, latency, cost per request, resilience to bad input, and human evaluation of utility. Record the baseline value before improvement and the test conditions so the result does not look accidental.
Reliability. Test empty, erroneous, and edge-case inputs, external service unavailability, and clear recovery. Do not send secrets and personal data to external models, verify facts, disclose AI usage, and separate demo results from overall performance.
Demo. Show a typical example, complex or erroneous input, tracing of key steps, measured result, and safe handoff to a human in case of uncertainty. Keep a backup local recording in case of a network error, if the rules allow.
Submission. Link every claimed benefit to a screen, measurement, source, or feedback. Clearly separate what works now from post-hackathon plans.

This strategy is not an official evaluation criterion. Its goal is to help turn the broad topic of AI agents, data and ML. The standard exclusion list does not contain Kazakhstan. into an honest, verifiable, and complete prototype.

Known risks and gaps

Identified risk: Take into account the requirements for novelty and public demonstration.. Next recommended step: Build an agent application with DataHub before 10.08.

The basic card fields are filled in, but before registering, you still need to check the exact hours, the current version of the rules, and any changes after the data collection date.

Practical go/no-go criterion: it is worth submitting if the team documentarily meets eligibility, has access to the required stack, has time to complete the verifiable scenario, and accepts financial and logistical risks. Participation should be paused if Kazakhstan's eligibility, submission rights, mandatory in-person presence, the cost of critical services, or the payout procedure for a significant prize are unclear.

Sources and level of reliability

Official event page.
Rules or additional source.
Recorded proof of eligibility: The standard exclusion list does not contain Kazakhstan..
The date and version of the documents may change; reopen both links before taking action.

Summary: Build with DataHub: The Agent Hackathon looks like an opportunity on the topic of "artificial intelligence, agents, and data" with the status Open and Kazakhstan's participation rated as Yes — according to the rules. Recommended action — Build an agent application with DataHub before 10.08. The decision should be made after addressing the listed gaps, not just based on the appeal of the topic or the prize headline.